>>> On 07.09.12 at 10:46, Huang Ying <ying.huang@xxxxxxxxx> wrote: > On Fri, 2012-09-07 at 08:12 +0100, Jan Beulich wrote: >> >>> On 07.09.12 at 09:02, Huang Ying <ying.huang@xxxxxxxxx> wrote: >> > On Fri, 2012-09-07 at 07:40 +0100, Jan Beulich wrote: >> >> At least for Xen Dom0 it is very relevant to set the _PAGE_IO bit when >> >> ioremap-ing (non-RAM) areas, as that flag suppresses the PFN -> MFN >> >> translation otherwise done in PV guests. (The offending commit was >> >> 81e88fdc432a1552401d6e91a984dcccce72b8dc.) >> > >> > If my understanding were correct, ghes_ioremap_pfn_nmi normally only map >> > RAM. >> >> But not RAM in the sense of what is seen as usable memory in the >> E820 map. What you're mapping here is reserved memory, and that >> is (from the perspective of the kernel and Xen) non-RAM. The fact >> that it is RAM from a machine perspective is manifested by the >> mapping not needing to be done UC or alike. > > I think the key is to keep these memory range map consistent between > each other. BIOS may map these memory cached. ACPI may mapped them as > cached (see acpi_os_ioremap) too. But that aspect is orthogonal to the patch here - all we need done for Xen is the proper setting of _PAGE_IO; cachability is being dealt with independently (and correctly I think). Jan -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html